Reading Disappearing Text:

advertisement
COPY OF:
Rayner, K., Liversedge, S.P., White, S.J., & Vergilino-Perez, D. (2003). Reading
disappearing text: Cognitive control of eye movements. Psychological Science, 14, 385-388.
NOT THE FINAL PUBLISHED VERSION
1
Reading Disappearing Text:
Cognitive Control of Eye Movements
Keith Rayner
University of Massachusetts, Amherst
Simon P. Liversedge, Sarah J. White
University of Durham
and
Dorine Vergilino-Perez
Université René Descartes Paris 5
Correspondence to:
Keith Rayner
Department of Psychology
University of Massachusetts
Amherst, MA 01003, USA
413-545-2175
rayner@psych.umass.edu
Running Head: Reading Disappearing Text
Word Count = 2499
2
Abstract
Readers read sentences containing high or low frequency target words under normal
reading conditions or disappearing text conditions (in which the word that was fixated
disappeared after 60 ms). Even though the fixated word had disappeared after 60 ms, there
was still a robust frequency effect wherein readers fixated longer on low frequency words
than on high frequency words. As such, the results are consistent with cognitive control
models of eye movement control and inconsistent with visual/oculomotor control models.
While the uptake of visual information is clearly important for reading, it is the cognitive
processes associated with understanding the fixated words that drive the eyes through the
text.
3
It is well-known that readers make a series of eye movements (called saccades)
separated by fixations (Rayner, 1998). It is during the eye fixations (which typically last
200-300 ms) that new information is acquired by the processing system and the mental
representation of what the text means is constructed. Eye movements are critical because
they mediate the complex sequence of cognitive processes that are involved in first extracting
the required visual information from the text and secondly the interpretation of that
information. The question we address in this article is: What determines when the eyes move
during reading? Three general types of models have been proposed to account for eye
movement control in reading: minimal control, visual/oculomotor control, and cognitive
control.
According to the minimal control view (Bouma & deVoogd, 1974; Kolers, 1976;
Suppes, 1994), there is no relationship between eye movements and linguistic processing of
the text: eye movements serve the function of bringing new information into the processing
system, but there is no relationship between how long a reader fixates on a particular word
and any linguistic property of that word. However, this view has been largely abandoned due
to numerous findings demonstrating that how long a reader looks at a word is influenced by
the ease or difficulty associated with processing that word (see Rayner, 1998, for a review).
Perhaps the major mitigating finding against the minimal control view is the word frequency
effect: readers look longer at low frequency words than at high frequency words (Inhoff &
Rayner, 1986; Rayner & Duffy, 1986). This effect is quite robust and has been demonstrated
many times (Rayner, 1998), and is one we utilize in the present experiment.
The visual/oculomotor control view (O'Regan, 1992; Yang & McConkie, 2001) holds
4
that the primary determinant of when to move the eyes is the visual/oculomotor processes
associated with reading. According to this view, when the eyes move is only indirectly
related to the mental activities associated with ongoing language processing. O'Regan (1992)
suggested that linguistic factors influence the duration of a single long fixation (300 ms or
longer), and only the second of two fixations when words are refixated (i.e., receive
additional fixations prior to the reader moving to another word). Yang and McConkie (2001)
likewise argued that cognitive processing activities only influence long fixations and
variations in fixation times have little relation to a reader=s current cognitive activity. Yang
and McConkie do not deny that there are linguistic influences on fixation times (such as the
word frequency effect). But, within their Competition-Interaction model most of the
variation in fixation time is due to physiological/visual processes that Ahave little relation to
the current cognitive activity@ (pg 3584). A basic feature of their model is that when
cognitive factors influence eye fixations it is by canceling (and reprogramming) a saccade (as
opposed to cognitive activities initiating a new saccade as in the cognitive control models).
According to the cognitive control view (Morrison, 1984; Reichle, Pollatsek, Fisher,
& Rayner, 1998; Reichle, Pollatsek, & Rayner, 1999; Reichle, Rayner, & Pollatsek, 2002;
Henderson & Ferreira, 1990), linguistic processing of the text largely determines when the
eyes move. For example, in the E-Z Reader model (Reichle et al., 1998), a familiarity check
that precedes lexical access serves as the signal for the initiation of an eye movement and
completion of lexical access is the signal for a shift of covert attention from the fixated word
to the next word in the text. This type of model can easily account for the fact that low
frequency words are fixated longer than high frequency words.
5
In the present study, we used a novel paradigm in which text literally disappeared
from before a reader's eyes (see Figure 1) to further determine what controls when the eyes
move in reading. Each time the eyes fixated on a word, it disappeared after 60 ms. When
readers moved their eyes to a new word, it too disappeared after 60 ms. An immediate
refixation on the disappearing word was not functional in that it did not result in that word
reappearing. With this paradigm, we compared fixation times on high and low frequency
words when readers read normal and disappearing text. Cognitive control models predict that
the frequency effect should emerge independent of whether readers read normal or
disappearing text. That is, since it is the mental activities associated with processing a given
word that drive the eyes through the text, readers should remain fixated on a low frequency
word longer than on a high frequency word, even though the word has disappeared. Since
there are variations of visual/oculomotor models, they might differ somewhat in the
prediction that they make about what should happen when reading disappearing text.
However, it seems that in general they should predict no difference between how long
readers look at high and low frequency words under disappearing text conditions. That is,
since the visual information needed for reading is gone after 60 ms, the eyes should move
after approximately the same amount of time regardless of whether readers fixate low or high
frequency target words.
Insert Figure 1 about here
Method
Participants. Eight undergraduate students at the University of Durham participated
in the experiment. They were all skilled readers with normal uncorrected vision, native
6
speakers of English, and naïve concerning the purpose of the experiment.
Materials and Design. There were 40 experimental sentences. For each sentence,
two versions were created. In one version, a high frequency target word was embedded in
the sentence; in the second version, a low frequency word occupied the same location as the
high frequency word. Either word fit naturally in the sentence. Some example sentences
with high and low frequency target words are shown in Table 1.
Insert Table 1 about here
Word frequencies were calculated using the CELEX English word form corpus
(Baayen, Piepenbrock, & Gulikers, 1995). The frequent words had a mean frequency of 105
per million (range: 18-970) and the low frequency words had a mean frequency of 1 (range:
0-7). Each sentence was one line of text (with no more than 80 character spaces per line).
The 40 experimental sentences were presented in two blocks (one in normal text format and
one in disappearing text format). The order of blocks was counterbalanced across
participants so that each of them received high or low frequency target words under the
normal or disappearing condition equally often. Five practice sentences were presented at
the beginning of each block and 8 comprehension questions were presented in each block;
these questions appeared on the video monitor and participants responded by pushing one
button for a 'yes' answer and another button for a 'no' answer.
Apparatus. The sentences were presented on a Phillips 21B582BH 21 inch monitor
(with a P22 phosphor with a decay to zero in less than 1.7 milliseconds) interfaced with a
Phillips Pentium III computer, which in turn was interfaced with a Fourward Technologies
Dual-Purkinje Generation 5.5 eye-tracker (with spatial resolution of less than 10 min of arc).
7
The sentences were displayed as white letters (in lower case except where capital letters
were appropriate) on a black background. Five characters subtended one degree of visual
angle. Viewing was binocular, but only the movements of the right eye were monitored.
Eye position was sampled every millisecond and the fixated word disappeared 60 ms after the
computer determined that the eye had fixated on the word1.
Procedure. Participants were instructed to read the sentences so that they understood
them, regardless of how they were presented on the monitor (i.e., normal text or disappearing
text). A bite bar and head restraint were used to minimize head movements. The initial
calibration procedure lasted approximately 5 minutes and calibration accuracy was checked
prior to every trial in the experiment. After reading each sentence, the participants pressed a
button to continue; the button press resulted in either a comprehension question or the
calibration pattern followed by the next sentence in the sequence.
Results
Although the disappearing text situation might appear to be somewhat bizarre, readers
could read quite easily. Indeed, their reading rate (265 wpm for disappearing text and 270
wpm for normal text) and comprehension level were the same for disappearing (94% correct)
and normal text (96% correct).
Consistent with standard practices in eye movement research (Rayner, 1998;
Liversedge & Findlay, 2000), we examined three measures of initial processing time for the
target words: single fixation duration (where readers made only one fixation on the target
word), first fixation duration (the duration of the first fixation on a word independent of the
number of fixations that were made on the word), and gaze duration (the sum of all fixations
on a word before moving the eyes off that word). Since gaze duration is typically considered
8
the best measure of processing time for a word, we will present those data first. For each of
the three measures, 2 (text: normal vs. disappearing) X 2 (frequency: high vs. low) analyses
of variance were carried out based on participant (F1) and item variability (F2).
Insert Table 2 about here
Table 2 shows the means for the three measures of processing time. For gaze
duration, there was a very robust effect of frequency as readers fixated longer on low
frequency words (331 ms) than on high frequency words (290 ms), F1(1,7) = 18.02, p < .005;
F2 (1,39) = 11.09, p < .005. However, there was no effect of text condition (Fs < 1) and no
interaction (Fs < 1).
The first fixation and single fixation duration data were consistent with the gaze
duration data. For single fixation duration, there was again a significant frequency effect, F1
(1,7) = 10.75, p < .05; F2 (1,39) = 22.02, p < .001, as the low frequency words (303 ms) were
fixated longer than high frequency words (267 ms), but neither the main effect of text (ps >
.15) nor the interaction (ps > .25) was significant. For first fixation duration, there was again
a robust frequency effect with longer fixations for low frequency words (295 ms) than high
frequency words (266 ms), F1 (1,7) = 13.06, p < .01; F2 (1,39) = 11.02, p < .005. Neither the
main effect of text, F1 (1,7) = 4.38, p > .07; F2 (1,39) = 3.82, p > .05, nor the interaction (Fs
< 1) was significant. The tendency for slightly longer first fixations for disappearing text was
offset by the fact that readers were much less likely (6.5%) to refixate on the target word with
disappearing text than with normal text (14%), F1 (1,7) = 9.0, p < .05; F2 (1,39) = 5.32, p <
.05. Of course, the differences in refixation tendency combined with the fixation times to
yield virtually equivalent means for normal and disappearing text in the gaze duration
measure.
Insert Figure 2 about here
Consistent with prior research (Rayner, 1995), the distribution for the normal reading
9
condition showed a clear rightward shift in the frequency distribution for low frequency
words relative to high frequency words2. The disappearing text condition likewise showed a
shift of the distribution (see Figure 2). Thus, it is the case that the difference in mean gaze
durations for high and low frequency words was not due to a number of very long gaze
durations as the visual/oculomotor control models would suggest.
Discussion
Our results demonstrate that it is the mental processes associated with understanding
the fixated word that determine when the eyes move in reading. Even though the fixated
word disappeared after 60 ms, readers continued fixating longer on the location of where the
word had been for low frequency words than for high frequency words. As such, the results
are consistent with cognitive control models, such as the E-Z Reader model (Reichle et al.,
1998), which posit that lexical processing determines when the eyes move. Furthermore, the
results are inconsistent with visual/oculomotor models, such as the Competition-Interaction
model of Yang and McConkie (2001). According to Yang and McConkie (2001), cognitive
processing activities mainly serve to delay the onset of the next saccade by cancelling and
reprogramming a new saccade (and they used abrupt onsets and offsets to demonstrate such
effects). Thus, the disappearance of the fixated word in our study should delay the onset of
the next saccade. Indeed, in the case of first fixation duration (and single fixation duration)
there was a 20 ms delay in the next saccade under disappearing text conditions. However,
this was offset by the fact that readers were much less likely to refixate on a word in the
disappearing text condition than in the normal text condition. This resulted in equivalent
gaze durations on the target words in the disappearing and normal text conditions.
Furthermore, the fact that the size of the frequency effect was equivalent across the
disappearing and normal text conditions is difficult for the visual/oculomotor control models
to explain, whereas the result is easily explainable within cognitive control models.
10
Not only are our results consistent with cognitive control models, they are also
consistent with prior research which demonstrated that when text was available for 50-60 ms
before a masking pattern appeared, reading was quite normal (Rayner, Inhoff, Morrison,
Slowiaczek, & Bertera, 1981; Slowiaczek & Rayner, 1987; Ishida & Ikeda, 1989). This does
not mean that the mental processes associated with reading are completed in 50-60 ms, but
rather that this is the time required for retinal stimulation to produce excitation in the visual
cortex, thus providing the visual information necessary for cognitive processing to occur.
This short neural transmission time ensures that visual information is delivered to the
cognitive system very rapidly thereby enabling readers to make short eye fixations, typically
lasting approximately 200-300 ms.
In conclusion, the experiment presented here provides clear support for cognitive
control models of eye movements in reading. While the uptake of visual information is
important and necessary for reading to occur, it is the mental operations on that information
that drive the eyes through the text.
11
References
Baayen, R.H., Piepenbrock, R., & Gulikers, L. (1995). The CELEX lexical
database. (CD-ROM). Philadelphia: Linguistic Data Consortium, University of
Pennsylvania.
Bouma, H, & deVoogd, A.H. (1974). On the control of eye saccades in reading.
Vision Research, 14, 273-284.
Henderson, J.M., & Ferreira, F. (1990). Effects of foveal processing difficulty on the
perceptual span in reading: Implications for attention and eye movement control. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 16, 417-429.
Inhoff, A.W., & Rayner, K. (1986). Parafoveal word processing during eye fixations
in reading: Effects of word frequency. Perception & Psychophysics, 40, 431-439.
Irwin, D.E., Yantis, S., & Jonides, J. (1983). Evidence against visual integration
across saccadic eye movements. Perception & Psychophysics, 34, 49-57.
Ishida, T., & Ikeda, M. (1989). Temporal properties of information extraction in
reading studied by a text-replacement technique. Journal of the Optical Society A: Optics
and Image Science, 6, 1624-1632.
Kolers, P.A. (1976). Buswell's discoveries. In R.A. Monty & J.W. Senders (Eds.),
Eye movements and psychological processes (pp. 371-395). Hillsdale, NJ: Erlbaum.
Liversedge, S.P., & Findlay, J.M. (2000). Saccadic eye movements and cognition.
Trends in Cognitive Science, 4, 6-14.
Morrison, R.E. (1984). Manipulation of stimulus onset delay in reading: Evidence for
parallel programming of saccades. Journal of Experimental Psychology: Human
Perception and Performance, 10, 667-682.
O'Regan, J.K. (1992). Optimal viewing position in words and the strategy-tactics
theory of eye movements in reading. In K. Rayner (Ed), Eye movements and visual
12
cognition: Scene perception and reading (pp. 333-354). New York: Springer-Verlag.
Rayner, K. (1995). Eye movements and cognitive processes in reading, visual search,
and scene perception. In J.M. Findlay, R. Walker, & R.W. Kentridge (Eds), Eye movement
research: Mechanisms, processes and applications (pp. 3-12). Amsterdam: North
Holland.
Rayner, K. (1998). Eye movements in reading and information processing: 20 years
of research. Psychological Bulletin, 124, 372-422.
Rayner, K., & Duffy, S.A. (1986). Lexical complexity and fixation times in reading:
Effects of word frequency, verb complexity, and lexical ambiguity. Memory & Cognition,
14, 191-201.
Rayner, K., Inhoff, A.W., Morrison, R., Slowiaczek, M.L., & Bertera, J.H. (1981).
Masking of foveal and parafoveal vision during eye fixations in reading. Journal of
Experimental Psychology: Human Perception and Performance, 7, 167-179.
Reichle, E.D., Pollatsek, A., Fisher, D.L., & Rayner, K. (1998). Toward a model of
eye movement control in reading. Psychological Review, 105, 125-157.
Reichle, E.D., Rayner, K., & Pollatsek, A. (1999). Eye movement control in reading:
Accounting for initial fixation locations and refixations within the E-Z Reader model. Vision
Research, 39, 4403-4411.
Reichle, E.D., Rayner, K., & Pollatsek, A. (in press). The E-Z Reader model of eye
movement control in reading: Comparisons to other models. Behavioral and Brain
Sciences.
Slowiaczek, M.L., & Rayner, K. (1987). Sequential masking during eye fixations in
reading. Bulletin of the Psychonomic Society, 25, 175-178.
Suppes, P. (1994). Stochastic models of reading. In J. Ygge & G. Lennerstrand
(Eds.), Eye movements in reading (pp. 349-364). Oxford: Pergamon Press.
13
Yang, S.N., & McConkie, G.W. (2001). Eye movements during reading: a theory of
saccade initiation times. Vision Research, 41, 3567-3585.
14
Acknowledgments
This research was conducted when the first and fourth authors were visitors at the
University of Durham. It was supported by National Institute of Health Grant HD26765 and
by a grant from The Leverhulme Trust (United Kingdom) which supported the first author as
a Leverhulme Visiting Professor at the University of Durham. The fourth author was
supported by a Post-Doctoral Fellowship from the Foundation Fyssen (France). The authors
thank Bob Metcalfe for programming assistance and John Findlay, John Henderson, Albrecht
Inhoff, and Bob Kentridge for their helpful comments.
Requests for reprints should be sent to Keith Rayner, Department of Psychology,
University of Massachusetts, Amherst, MA 01003, USA; email: rayner@psych.umass.edu
15
Footnotes
1. We performed a shutter test (Irwin, Yantis, & Jonides, 1983) to ensure that
phosphor persistence did not enable readers to see a fading image of the fixated
word. Four participants, who did not participate in the main experiment, were
asked to perform a lexical decision task for words and non-words presented
centrally on the video monitor used in the main experiment. A fixation cross was
initially presented followed by a letter string which was presented for 60 ms. A
shutter device was programmed to close simultaneously with the presentation of
the letter string and to open coincident with the offset of this stimulus. If there
were a fading image due to phosphor persistence, the participants should have
been able to accurately respond in the task. However, all of them found the task
virtually impossible (they were all convinced that nothing had been presented
prior to the shutter opening, except for the fixation cross) and performance for all
four participants was completely at chance. Thus, it is clear that phosphor
persistence was not contributing to the results.
2. The frequency distributions for single fixation and first fixation duration likewise
showed a rightward shift of the distribution for low frequency words (for both the
normal and disappearing text conditions).
16
Table 1. Example sentences with the high and low frequency target word
1. Sam wore the horrid coat though his pretty/demure girlfriend complained.
2. He found the secret manuscript inside the little/sturdy farmhouse on the hill.
3. The clumsy volunteers asked the random/nimble gentleman to help carry the table.
4. Yesterday the office supervisor moaned about the broken/snazzy equipment upstairs.
5. A proper collection scheme boosted the annual/frugal donations to the charity.
Note: all target words were 6 letters long.
17
Table 2. Single Fixation Duration (SFD), First fixation duration (FFD), and Gaze Duration
(GD) means (in ms) as a function of text condition and frequency.
SFD
FFD
GD
High Frequency
257
256
289
Low Frequency
298
288
328
High Frequency
277
276
291
Low Frequency
307
301
333
Normal Text
Disappearing Text
18
Figure Captions
Figure 1. Example of disappearing text. When the reader fixated on the word church it was
visible for 60 ms. Then it disappeared until the reader made a fixation on a new word. Thus,
an immediate refixation on church did not result in it reappearing. However, if the reader
fixated on underwent (which disappeared after 60 ms) and then went back to church, it would
again be visible for 60 ms before disappearing. Each word in the sentence that the reader
fixated (beginning with the first word in the sentence) likewise disappeared after 60 ms of
each fixation. The asterisk indicates fixation location.
Figure 2. Frequency distribution of gaze durations for low and high frequency words in the
disappearing text condition (35 ms bins).
19
The really marvellous church underwent shoddy
*
The really marvellous
underwent shoddy
*
The really marvellous church underwent shoddy
*
The really marvellous church
shoddy
*
20
New Fixation
After 60 ms
New Fixation
After 60 ms
35
Frequent
30
Frequency (%)
Infrequent
25
20
15
10
5
0
80
185
290
395
500
Gaze Duration
21
605
710
Download